Intelligent Analysis And Fault Diagnosis Of Power Disnatching Based On Alarm Signal Text Mining

The power dispatching system will receive a large number of alarm signals when the power system breaks down.If the dispatchers cannot make a decision promptly,the fault may expand.The alarm signal of the power dispatching system is unstructured Chinese short text.which contains a large number of power system professional vocabulary,and is mixed with numbers and special symbols.In order to extract valuable information from the alarm signal,applying nature lan guage processing method to mining the information of the alarm signal have become a key step.In this paper,we propose an intelligent analysis and fault diagnosis method for power dispatching,which is based on alarm signal text mining.This method integrates Chinese nat-ural language processing technology,big data mining technology and artificial intelligence technology to realize intelligent fault diagnosis and is applied in a provincial power dispatch-ing center.The main contribution of this paper includes:1)The power dispatching fault diagnosis system based on alarm signal text mining.The whole process is divided into two stages:processing the alarm signal text and fault diagnosis.At the first stage,an entity dictionary is construct,and a method based on Hidden Markov Model is applied to do word segmentation to the alarm signal text.Then after removing the stop word,the system adopts the Vector Space Model(VSM)to transform each word into a vector.During the second stage,the system first reads the real-time alarm signals.Then a Support Vector Machine(SVM)classifier is employed to distinguish the failure signal.If there is a failure signal then the system uses k-means clustering algorithm to present faults of higher probability for the dispatcher to refer to.The actual alarm signal of a provincial power dispatching center is used as an example to verify the effectiveness of the method.This method does not depend on the structural information of the power system and the protection action logic.The diagnosis speed is fast,and the diagnosis result is accurate and reliable.2)Realizin g the management of power dispatching fault templates.Based on the former aspect,an alarm signal template management system is constructed to realize efficient storage,fast search and timely update of alarm signals.The common alarm signal keywords are formed by extracting the text feature words from the alarm signal.Based on the common alarm signal key.words,an improved prefix tree is formed to construct a general fault template,and the structural features of the improved prefix tree are used to realize the management of the alarm signal template.3)Develop a power dispatching fault diagnosis system.Based on the above two aspects,combined with the actual engineering application of the power grid,we design and develop a fault diagnosis system for power dispatching.The system has three modules:dispatching knowledge base,real-time fault diagnosis and fault template management.It has been de-ployed and operated normally in a provincial power dispatching center,providing compre-hensive support for dispatchers.